Millimeter wave radar ranging value correction method considering multi-source error

Abstract

The invention discloses a tunnel millimeter wave radar ranging value correction method considering multi-source errors. The method comprises the following steps: acquiring a radar ranging value and a meteorological observation value; inputting the radar ranging value and the meteorological observation value into the trained radar ranging value correction model, and outputting a corrected millimeter-level tunnel radar ranging value; wherein the training of the radar ranging value correction model comprises the steps of obtaining results of a data acquisition experiment, including a radar ranging value, a meteorological observation value and a laser observation value, and constructing a data set after preprocessing; constructing a radar ranging value correction model, namely a mixed model of LSTM and Bi-LSTM; and training a radar ranging value correction model by using the data set, and outputting the trained radar ranging value correction model.

Description

Technical Field

[0001] The present invention belongs to the field of surveying and mapping technology, and in particular relates to a tunnel millimeter wave radar ranging value correction method taking into account multi-source errors. Background Art

[0002] Millimeter-wave radar ranging technology, with its advantages of high accuracy and real-time performance, is widely used for deformation monitoring in tunnels, such as railway and highway tunnels. However, the ranging accuracy of millimeter-wave radar in tunnels still faces a number of challenges. Achieving high-precision monitoring and early warning using millimeter-wave radar in tunnels requires obtaining highly accurate millimeter-wave radar ranging values. However, when using millimeter-wave radar for ranging in tunnels, ranging values ​​are susceptible to factors such as meteorological errors (such as temperature and relative humidity), distance delay errors, and electronic component delay errors. Summary of the Invention

[0003] In order to solve the problem in the prior art that when using millimeter-wave radar for ranging in a tunnel environment, the ranging value is easily affected by factors such as meteorological errors, distance delay errors, and electronic component delay errors, the present invention proposes a millimeter-wave radar ranging value correction method that takes into account multi-source errors, so that the millimeter-wave radar can still maintain high-precision monitoring capabilities under complex environmental conditions.

[0004] According to one aspect of the present invention, a method for correcting a tunnel millimeter-wave radar ranging value taking into account multi-source errors is provided, comprising:

[0005] Collect radar ranging values ​​and meteorological observation values;

[0006] The radar ranging value and the meteorological observation value are input into the trained radar ranging value correction model, and the corrected millimeter-level tunnel radar ranging value is output;

[0007] The training of the radar ranging value correction model includes:

[0008] Obtain the results of the data collection experiment, including radar ranging values, meteorological observation values, and laser observation values, and construct a data set after preprocessing;

[0009] A radar ranging value correction model was constructed, consisting of a hybrid LSTM and Bi-LSTM layer, a Dropout layer, and three fully connected layers. The connection structure is: LSTM and Bi-LSTM hybrid layer - Dropout layer - first fully connected layer - second fully connected layer - third fully connected layer. Raw data is input into the hybrid LSTM and Bi-LSTM layer for processing. High-dimensional features are processed through subsequent structures, and finally the third fully connected layer outputs the corrected millimeter-level tunnel radar ranging value.

[0010] Preset training parameters, input the radar ranging values ​​and meteorological observation values ​​in the dataset into the radar ranging value correction model, output the radar correction value, calculate the error between the radar correction value and the corresponding laser observation value in the dataset, adjust the model parameters through the back propagation algorithm, iterate the training, and output the trained radar ranging value correction model.

[0011] As a further implementation scheme, radar ranging values ​​and meteorological data are obtained by a tunnel millimeter wave radar ranging system, including a millimeter wave radar board and a meteorological collection device; radar ranging values ​​are collected by the millimeter wave radar board, by sending millimeter waves to the radar target and receiving the millimeter waves reflected by the target, and obtaining radar ranging values ​​based on the time difference between sending and receiving; the meteorological observation values ​​are collected by the meteorological parameter collection device

[0012] As a further embodiment, meteorological observations are measured values ​​of environmental parameters in the tunnel, including but not limited to temperature, relative humidity, air pressure, wind speed, light pressure and PM 2.5 .

[0013] As a further implementation plan, the experimental equipment for the data acquisition experiment includes: a millimeter-wave radar board, a meteorological parameter acquisition device, a laser tracker and a corner reflector. The corner reflector serves as a radar target, the millimeter-wave radar board and the meteorological parameter acquisition device are used to collect radar ranging values ​​and meteorological observation values ​​respectively, and the laser tracker is used to collect laser observation values.

[0014] As a further implementation scheme, the data collection experiment is designed as follows:

[0015] Fix the millimeter-wave radar board and meteorological parameter collection device at the measurement point, fix the radar emission direction, install a laser tracker so that the laser emission direction and distance are consistent with the direction and distance of the radar, set the collection distance range, and divide the collection distance range into target value segments of equal distance. The corner reflector is used as the radar target and installed in the radar emission direction. The distance from the millimeter-wave radar board is changed according to different target value segments. For each target value segment, experimental data is collected within the preset fixed collection time.

[0016] As a further implementation plan, the collection of experimental data adopts a time-divided repeated observation scheme, dividing the preset fixed collection time into time periods. For each target value segment, the experimental data are repeatedly collected in different time periods. During the experiment, a laser tracker is used to collect laser observation values ​​from the measurement point to the millimeter-wave radar target.

[0017] As a further implementation plan, the radar LSTM and Bi-LSTM mixed layer in the radar ranging value correction model contains two LSTM layers and two Bi-LSTM layers, and the connection structure is: first Bi-LSTM layer—first LSTM layer—second LSTM layer—second Bi-LSTM layer.

[0018] The original data is input into the first Bi-LSTM layer, and is processed in sequence by the first Bi-LSTM layer, the first LSTM layer, the second LSTM layer, and the second Bi-LSTM layer. The high-dimensional features are output in the second Bi-LSTM layer. The Dropout layer discards the high-dimensional features according to the random rate. Finally, the discarded high-dimensional features are processed in sequence by the first fully connected layer, the second fully connected layer, and the third fully connected layer. The extracted high-dimensional features are mapped to the calibrated millimeter-level tunnel radar ranging values.

[0019] As a further implementation plan, the radar ranging value correction model works as follows: the first Bi-LSTM layer captures the long-term dependency between the radar ranging values ​​and the contextual information of meteorological parameters in the original data sequence. The first LSTM layer uses memory units and gating mechanisms to prevent overfitting and learn to preserve long-term dependencies. The second LSTM layer deeply extracts sequence features, further learning and capturing the complex patterns of long-term dependencies between sequences. Finally, the second Bi-LSTM layer receives the results of the second LSTM layer and obtains high-dimensional features with rich feature representation by integrating forward and backward information.

[0020] The results of the second Bi-LSTM layer are input to the Dropout layer. The dropout rate is set to randomly discard the output of the Dropout layer neurons. The high-dimensional features after the discarding process are linearly transformed and fused by the first fully connected layer. Then, the second fully connected layer reduces the dimension and fuses the features. Finally, the third fully connected layer maps them to a one-dimensional regression prediction value as the corrected millimeter-level tunnel radar ranging value.

[0021] According to another aspect of the present invention, a tunnel millimeter wave radar ranging value correction device taking into account multi-source errors is provided, which is used to implement a tunnel millimeter wave radar ranging value correction method taking into account multi-source errors, including:

[0022] Tunnel millimeter-wave radar ranging module, used to collect radar ranging values ​​and meteorological observation values;

[0023] The radar ranging value correction module is used to input the collected radar ranging values ​​and meteorological observation values ​​into the trained radar ranging value correction model and output the corrected millimeter-level tunnel radar ranging value; wherein the training of the radar ranging value correction model includes:

[0024] Obtain the results of the data collection experiment, including radar ranging values, meteorological observation values, and laser observation values, and construct a data set after preprocessing;

[0025] A radar ranging value correction model was constructed, consisting of a mixed LSTM and Bi-LSTM layer, a Dropout layer, and three fully connected layers. The connection structure was: LSTM and Bi-LSTM mixed layer—Dropout layer—first fully connected layer—second fully connected layer—third fully connected layer. The LSTM and Bi-LSTM mixed layer contained two LSTM layers and two Bi-LSTM layers, with a connection structure: first Bi-LSTM layer—first LSTM layer—second LSTM layer—second Bi-LSTM layer. Raw data was input into the mixed LSTM and Bi-LSTM layers, which then processed and output high-dimensional features. The high-dimensional features were then processed by the subsequent structure, and finally the corrected millimeter-level tunnel radar ranging value was output in the third fully connected layer.

[0026] Preset training parameters, input the radar ranging values ​​and meteorological observation values ​​in the dataset into the radar ranging value correction model, output the radar correction value, calculate the error between the radar correction value and the corresponding laser observation value in the dataset, adjust the model parameters through the back propagation algorithm, iterate the training, and output the trained radar ranging value correction model.

[0027] As a further implementation method, the correction equipment also includes a millimeter-wave radar board and a meteorological parameter acquisition device, which are respectively connected to the tunnel millimeter-wave radar ranging module, wherein the millimeter-wave radar board sends millimeter waves to the radar target and receives millimeter waves reflected by the target, and obtains the radar ranging value based on the time difference between sending and receiving, and the meteorological parameter acquisition device collects meteorological observation values.

[0028] As a further embodiment, it also includes: a laser tracker and a corner reflector, wherein the corner reflector serves as a radar target for the millimeter wave radar board and the meteorological parameter acquisition device to respectively collect radar ranging values ​​and meteorological observation values, and the laser tracker is used to collect laser observation values.

[0029] According to another aspect of the present invention, an electronic device is provided, comprising a memory and a processor, wherein the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute a method for correcting tunnel millimeter-wave radar ranging values ​​that takes into account multi-source errors.

[0030] According to another aspect of the present invention, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to execute a method for correcting tunnel millimeter-wave radar ranging values ​​taking into account multi-source errors.

[0031] Compared with the existing technology, the beneficial effect of the present invention lies in: based on the problem that current millimeter-wave radars are easily affected by factors such as meteorological errors (such as temperature, relative humidity, etc.), distance delay errors, and electronic component delay errors in tunnels, the present invention proposes a tunnel millimeter-wave radar ranging value correction method that takes into account multi-source errors. This method can improve the ranging accuracy of millimeter-wave radars from the centimeter level to the millimeter level or even the submillimeter level, and can play an extremely important role in tunnel environments with extremely high precision requirements. At the same time, the equipment is economical and affordable, and can greatly expand the market to provide convenience to society. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction will be given below to the drawings used in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0033] Figure 1 1 is a flow chart of a method for correcting tunnel millimeter-wave radar ranging values ​​taking into account multi-source errors in an embodiment of the present invention;

[0034] Figure 2 Schematic diagram of the process of radar ranging value correction model training in an embodiment of the present invention;

[0035] Figure 3 Schematic diagram of the structure of the radar ranging value correction model in an embodiment of the present invention;

[0036] Figure 4 A schematic diagram of a radar ranging experiment design according to an embodiment of the present invention;

[0037] Figure 5 Schematic diagram of the data preprocessing process in an embodiment of the present invention;

[0038] Figure 6 Schematic diagram of training of radar ranging value correction model in an embodiment of the present invention;

[0039] Figure 7 Schematic diagram of a tunnel millimeter-wave radar ranging value correction device taking into account multi-source errors in an embodiment of the present invention;

[0040] Figure 8 Schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] It should be noted that:

[0042] The terms "including" and "having" and any variations thereof in the description and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions, for example, a process, method, system, product or apparatus that includes a series of steps or units is not necessarily limited to the steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0043] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices. The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily need to be executed in the order described. For example, some operations / steps may be further decomposed, while others may be combined or partially combined, so the actual execution order may vary depending on the actual situation.

[0044] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, the technical features in the various embodiments or single embodiments provided by the present invention are arbitrarily combined with each other to form a new technical solution. This combination is not restricted by the sequence of steps and / or structural composition mode, but must be based on the ability of ordinary technicians in this field to implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that this combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0045] like Figure 1 As shown, Figure 1 : This is a flow chart of a method for correcting tunnel millimeter-wave radar ranging values ​​taking into account multi-source errors in an embodiment of the present invention, including:

[0046] Collect radar ranging values ​​and meteorological observation values;

[0047] The radar ranging value and the meteorological observation value are input into the trained radar ranging value correction model, and the corrected millimeter-level tunnel radar ranging value is output;

[0048] Among them, such as Figure 3 As shown, the training of the radar ranging value correction model includes:

[0049] Obtain the results of the data collection experiment, including radar ranging values, meteorological observation values, and laser observation values, and construct a data set after preprocessing;

[0050] A radar ranging value correction model was constructed, consisting of a hybrid LSTM and Bi-LSTM layer, a Dropout layer, and three fully connected layers. The connection structure is: LSTM and Bi-LSTM hybrid layer - Dropout layer - first fully connected layer - second fully connected layer - third fully connected layer. Raw data is input into the hybrid LSTM and Bi-LSTM layer for processing. High-dimensional features are processed through subsequent structures, and finally the third fully connected layer outputs the corrected millimeter-level tunnel radar ranging value.

[0051] Preset training parameters, input the radar ranging values ​​and meteorological observation values ​​in the dataset into the radar ranging value correction model, output the radar correction value, calculate the error between the radar correction value and the corresponding laser observation value in the dataset, adjust the model parameters through the back propagation algorithm, iterate the training, and output the trained radar ranging value correction model.

[0052] Furthermore, radar ranging values ​​and meteorological data are obtained by a tunnel millimeter-wave radar ranging system, which includes a millimeter-wave radar board and a meteorological collection device; the radar ranging values ​​are collected by the millimeter-wave radar board by sending millimeter waves to the radar target and receiving millimeter waves reflected by the target, and the radar ranging values ​​are obtained according to the time difference between sending and receiving; the meteorological observation values ​​are collected by the meteorological parameter collection device.

[0053] Furthermore, meteorological observation values ​​are measured values ​​of environmental parameters in the tunnel, including but not limited to temperature, relative humidity, air pressure, wind speed, light pressure and PM 2.5 .

[0054] Furthermore, the experimental equipment for the data acquisition experiment includes: a millimeter-wave radar board, a meteorological parameter acquisition device, a laser tracker and a corner reflector. The corner reflector serves as a radar target, the millimeter-wave radar board and the meteorological parameter acquisition device are used to collect radar ranging values ​​and meteorological observation values ​​respectively, and the laser tracker is used to collect laser observation values.

[0055] Furthermore, the data collection experiment is designed as follows:

[0056] Fix the millimeter-wave radar board and meteorological parameter collection device at the measurement point, fix the radar emission direction, install a laser tracker so that the laser emission direction and distance are consistent with the direction and distance of the radar, set the collection distance range, and divide the collection distance range into target value segments of equal distance. The corner reflector is used as the radar target and installed in the radar emission direction. The distance from the millimeter-wave radar board is changed according to different target value segments. For each target value segment, experimental data is collected within the preset fixed collection time.

[0057] Specifically, the corner reflector is determined to be the millimeter-wave radar target, and it is fixed on the platform to ensure that it has good signal beam reflection. The millimeter-wave radar board is fixed on the measurement platform to ensure that it is not affected by the environment, so that the millimeter-wave radar board is in a stationary state and the measurement error is reduced.

[0058] Furthermore, the experimental data are collected using a time-divided repeated observation scheme, which divides the preset fixed collection time into time periods. For each target value segment, the experimental data are repeatedly collected in different time periods. During the experiment, a laser tracker is used to collect laser observation values ​​from the measurement point to the millimeter-wave radar target.

[0059] Specifically, a laser tracker must be used in each target value segment to ensure that the true observation value is obtained in three different time periods divided by a preset fixed acquisition time. Each time period must be measured at least three times to prevent accidental errors.

[0060] Specifically, the meteorological data, laser tracker data, and radar ranging value data are preprocessed and divided into training set, test set, and validation set in a ratio of 7:2:1.

[0061] Furthermore, the LSTM and Bi-LSTM hybrid layer in the radar ranging value correction model contains two LSTM layers and two Bi-LSTM layers, and the connection structure is: first Bi-LSTM layer - first LSTM layer - second LSTM layer - second Bi-LSTM layer;

[0062] The original data is input into the first Bi-LSTM layer, and is processed in sequence by the first Bi-LSTM layer, the first LSTM layer, the second LSTM layer, and the second Bi-LSTM layer. The high-dimensional features are output in the second Bi-LSTM layer. The Dropout layer discards the high-dimensional features according to the random rate. Finally, the discarded high-dimensional features are processed in sequence by the first fully connected layer, the second fully connected layer, and the third fully connected layer. The extracted high-dimensional features are mapped to the calibrated millimeter-level tunnel radar ranging values.

[0063] Furthermore, the radar ranging value correction model works as follows: the first Bi-LSTM layer captures the long-term dependency between the radar ranging values ​​and the contextual information of meteorological parameters in the original data sequence. The first LSTM layer uses memory units and gating mechanisms to prevent overfitting and learn to preserve long-term dependencies. The second LSTM layer deeply extracts sequence features, further learning and capturing the complex patterns of long-term dependencies between sequences. Finally, the second Bi-LSTM layer receives the results of the second LSTM layer and obtains high-dimensional features with rich feature representation by integrating forward and backward information.

[0064] The results of the second Bi-LSTM layer are input to the Dropout layer. The dropout rate is set to randomly discard the output of the Dropout layer neurons. The high-dimensional features after the discarding process are linearly transformed and fused by the first fully connected layer. Then, the second fully connected layer reduces the dimension and fuses the features. Finally, the third fully connected layer maps them to a one-dimensional regression prediction value as the corrected millimeter-level tunnel radar ranging value.

[0065] Specifically, if Figure 3 As shown in the figure, the working steps of the radar ranging value correction model are as follows:

[0066] The first Bi-LSTM layer captures the long-term dependencies between radar ranging values ​​and contextual information about meteorological parameters in the original data sequence. The first LSTM layer uses memory cells and gating mechanisms to prevent overfitting and learn to preserve long-term dependencies. The second LSTM layer deeply extracts sequence features, further learning and capturing the complex patterns of long-term dependencies between sequences. Finally, the second Bi-LSTM layer receives the results of the second LSTM layer and integrates forward and backward information to obtain high-dimensional features with rich feature representation.

[0067] The results of the second Bi-LSTM layer are input to the Dropout layer. The dropout rate is set to randomly discard the output of the Dropout layer neurons. The high-dimensional features after the discarding process are linearly transformed and fused by the first fully connected layer. Then, the second fully connected layer reduces the dimension and fuses the features. Finally, the third fully connected layer maps them to a one-dimensional regression prediction value as the corrected millimeter-level tunnel radar ranging value.

[0068] The first Bi-LSTM (bidirectional long short-term memory network) layer receives the original 8-dimensional raw data (hours, PM25, temperature, humidity, wind speed, air pressure, light, and ranging values) and processes the sequence data bidirectionally, capturing contextual information and mining long-term dependencies, which is suitable for time series data. The first LSTM (long short-term memory network) layer further processes the sequence data, using memory units and gating mechanisms to solve the RNN gradient problem and learn to preserve long-term dependencies. The second LSTM layer deeply extracts features, learns sequence relationships, and captures complex patterns. The third Bi-LSTM layer processes the data again bidirectionally, integrating forward and backward information and enriching feature representation.

[0069] The dropout layer (30% dropout rate) receives the output of the second Bi-LSTM layer and randomly drops neuron outputs with a probability of 30% to prevent overfitting and enhance generalization. The first fully connected layer (256 neurons) performs linear transformation and fusion on features, mapping high-dimensional features to a new space and learning complex feature combinations. The second fully connected layer (128 neurons) reduces dimensionality and fuses features, refining them for the final prediction. The third fully connected layer (1 neuron) maps the extracted features into a one-dimensional regression prediction value for the final prediction result.

[0070] A hybrid LSTM and Bi-LSTM model is used for training. This hybrid model effectively captures long-term dependencies and bidirectional information in time series data, enabling precise compensation and correction of ranging values. This significantly improves the ranging accuracy of millimeter-wave radar in tunnel environments, enabling it to maintain high-precision monitoring capabilities even in complex environments.

[0071] Furthermore, the model training parameters include:

[0072] The maximum number of training epochs, initial learning rate, learning rate decay factor, validation frequency, data shuffling period, and learning rate decay period.

[0073] Specifically, we train the LSTM and Bi-LSTM hybrid models and test them using the test set data. The operations include:

[0074] 1) Start: The starting point of the process.

[0075] 2) Determine the network structure and related parameters: Determine the architecture of the neural network, such as the number of layers, number of neurons, etc., as well as hyperparameters such as learning rate.

[0076] 3) Hybrid model parameter initialization: assign initial values ​​to the parameters in the model, such as weights and biases.

[0077] 4) Build a hybrid model: Build a model based on the determined structure and initialization parameters.

[0078] 5) Forward propagation calculation: The training set data is normalized and flattened before being input into the model to calculate the output results.

[0079] 6) Back propagation and error calculation: Calculate the error based on the forward propagation result and the true value, and adjust the model parameters through the back propagation algorithm.

[0080] 7) Model validation calculation: Use the validation set (normalized and flattened) to evaluate model performance.

[0081] 8) Judgment conditions: Check whether the gradient error meets the requirements, whether the number of iterations reaches the maximum, or whether the loss is improved after multiple epochs. If so, proceed to the next step. Otherwise, increase the number of iterations and the epoch by 1 and return to continue training.

[0082] 9) Is the training finished?: Further confirm whether the training is terminated. If not, return to continue training.

[0083] 10) Network model performance test: Use the test set (normalized and flattened) to evaluate the trained model.

[0084] 11) Calculate performance evaluation indicators: calculate mean absolute error (MAE), mean absolute percentage error (MAPE) and root mean square error (RMSE).

[0085] 12) Millimeter-wave radar ranging prediction value output: Output model’s prediction result for millimeter-wave radar ranging.

[0086] 13) End: The process ends.

[0087] This invention is based on the existing components of conventional millimeter-wave radar. A complete millimeter-wave radar system includes transmitting (TX) and receiving (RX) radio frequency (RF) components, as well as analog components such as a clock, and digital components such as an analog-to-digital converter (ADC), a microcontroller (MCU), and a digital signal processor (DSP). In radar systems, distance measurement fundamentally involves reflections from electromagnetic signals that are blocked by objects in their path. The frequency of the signal used in FMCW radar systems increases linearly over time, and this invention utilizes FMCW linear frequency modulated continuous pulses. The transmitted signal and the echo signal are mixed to generate an intermediate frequency signal, which is then subjected to a fast Fourier transform to obtain target distance information.

[0088] Based on the problem that millimeter-wave radars are currently susceptible to meteorological errors (such as temperature and relative humidity), distance delay errors, and electronic component delay errors in tunnels, this invention proposes a tunnel millimeter-wave radar ranging value correction method that takes into account multi-source errors. This method can improve the ranging accuracy of millimeter-wave radars from the centimeter level to the millimeter level or even the submillimeter level, and can play an extremely important role in tunnel environments with extremely high precision requirements. At the same time, the equipment is economical and affordable, and can greatly expand the market and provide convenience to society.

[0089] An example of tunnel millimeter-wave radar ranging value correction taking into account multi-source errors includes:

[0090] Step 1: Identify the octagonal corner reflector as the millimeter-wave radar target and secure it to the platform to ensure good signal beam reflection. Secure the millimeter-wave radar board to the measurement platform to ensure it is not affected by the environment. Keep the millimeter-wave radar board stationary to reduce measurement errors.

[0091] Step 2: The millimeter wave radar is tested according to the designed ranging experiment;

[0092] Step 3: During the radar ranging experiment, the measurement time of each distance segment is 24 hours, and a micro-meteorological instrument is used to obtain meteorological data, and a laser tracker is used to obtain the actual observation value of each distance segment;

[0093] Step 4: Process the meteorological data, laser tracker data, and radar ranging data, and divide them into training set, test set, and validation set in a ratio of 7:2:1;

[0094] Step 5: Use the LSTM and Bi-LSTM hybrid model for training and test it using the test set data;

[0095] Step 6: After using the prediction model to predict the test set data, the test set data is corrected according to the linear fitting method in existing literature, and the correction effects of the two are compared and analyzed.

[0096] In this method, the experimental equipment mainly includes the millimeter wave radar board IWR6843ISK, Leica AT960 laser tracker, micro-meteorological instrument and octagonal corner reflector. The millimeter wave radar board combines hardware equipment, including the millimeter wave radar board, USB data cable, fixed platform, and computer, to achieve fast and accurate measurement. The millimeter wave radar board uses the industrial version of IWR6843ISK, which contains a 60 GHz millimeter wave radar transceiver. It is an integrated single-chip millimeter wave sensor based on FMCW radar technology and can operate in the 60 GHz to 64 GHz frequency band; the micro-meteorological instrument can integrate temperature, relative humidity, air pressure, wind speed, light intensity, PM 2.5 , wind direction, rainfall, etc. It can work all-weather and is not affected by severe weather such as heavy rain and frost; when the radar electromagnetic wave scans the corner reflector, the electromagnetic wave will be refracted and amplified on the metal corner, generating a strong echo signal, and a strong echo target will appear on the radar screen for subsequent data processing; the Leica AT960 single-point measurement accuracy can reach micron level, which is used to provide real observation values ​​between the radar and the octagonal corner reflector.

[0097] The above mainly includes data collection, data preprocessing, distance value prediction model establishment, distance value prediction correction and comparative experiments, a total of five steps, corresponding to steps two, three, four, five and six respectively, specifically:

[0098] Data acquisition, radar ranging experiment design such as Figure 4As shown, because high-precision millimeter-wave radar ranging is generally only applicable to short distances, the correction distance is set here to 5-21 meters. To determine the relationship between radar ranging values, climate parameters (such as temperature, humidity, and air pressure) and propagation distance, this experiment was designed to begin with measurements at a distance of 5 meters from the radar and continue at 2-meter intervals until 21 meters, for a total of nine target value segments. Each target value segment was measured for nearly 24 hours. Tunnel meteorological fluctuations are relatively small within a small range, and short-term variations are also relatively small. To better capture the relationship between radar ranging values ​​and parameters such as temperature and humidity, the data acquisition frequency of the micro-meteorological instrument was set to 10 Hz, recording data every 10 seconds. The actual observations between the radar and the target were obtained using a Leica AT960 laser tracker. To enhance the reliability of the actual observations, a time-segmented repeated observation scheme was used. Range measurements were performed on the octagonal corner reflector within the same distance segment at three different time periods throughout the day, with multiple repetitions for each measurement. Through triple error elimination and mean processing, the final true observation value of the target value segment of the octagonal corner reflector is ensured to be highly reliable.

[0099] Data preprocessing. The data preprocessing process is as follows Figure 5 As shown in the figure, the collected meteorological data, laser tracker data and radar ranging value data were processed by outlier removal and triple mean error removal, and then made into a data set, which was divided into training set, test set and validation set in the ratio of 7:2:1.

[0100] The distance value prediction model is established. The present invention adopts LSTM and Bi-LSTM hybrid model for training. The training process is as follows Figure 6 The model combines Bi-LSTM and LSTM layers, uses Dropout and fully connected layers, and ultimately outputs millimeter-wave radar ranging predictions through a regression layer. The key parameters for model training used in this invention are shown below.

[0101] Table 1 Key parameters for model training

[0102]

[0103] After training the prediction model, the test set data is used for prediction and analyzed based on indicators such as mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE).

[0104] The ranging value prediction results obtained by the ranging value prediction model are shown in Table 2. The two columns of true observations and radar ranging mean values, as well as the difference between them, show that before using the proposed method for correction, millimeter-wave radar ranging values ​​have large errors due to factors such as electronic component delay, distance delay, and meteorological factors. The maximum error is 100.74 mm, the minimum error is 76.66 mm, and the remaining errors are between 90 and 100 mm, resulting in low accuracy, only within the centimeter level. Using the proposed method for regression prediction of millimeter-wave radar ranging values, the predicted values ​​are subjected to three-fold mean error elimination and average processing. The last column clearly shows that the accuracy has been improved from the original centimeter level to the millimeter level, and even to the submillimeter level.

[0105] The performance evaluation results are shown in Table 3. Starting with the MAPE, a vertical comparison reveals that, with the exception of the 700 cm, 1500 cm, and 1700 cm segments, the MAPEs are relatively large, at 0.235740, 0.098102, and 0.086621, respectively. The MAPEs for the remaining target segments are relatively small. A horizontal comparison reveals that the MAE and RMSE for the 700 cm, 1500 cm, and 1700 cm segments are also relatively large, exceeding 10 mm. However, judging by the mean predictions, the accuracy of the predictions for the 700 cm, 900 cm, 1500 cm, 1700 cm, and 2100 cm segments reaches submillimeter accuracy, while the accuracy for the 500 cm, 1300 cm, and 1900 cm segments reaches 1 mm. The worst accuracy is 1.75 mm for the 500 cm segment.

[0106] Table 2 Schematic diagram of ranging value prediction results

[0107]

[0108] in, It represents the error between the mean value of the radar range measurement and the true observation value. The calculation formula is: , superscript Indicates the sequence number of the target value segment; It represents the error between the mean of the regression prediction value obtained by the ranging value prediction model and the true observation value. The calculation formula is: ,

[0109] Table 3 Schematic table of performance evaluation indicators

[0110]

[0111] Comparative experiment. After using the prediction model to predict the test set data, the test set data was corrected according to the linear fitting method in existing literature, and the correction effects of the two were compared and analyzed.

[0112] After using the prediction model to predict the test data, the test data was corrected using a linear fitting method from existing literature. The results are shown in Table 4. After the linear fitting method corrected the millimeter-wave radar ranging value, its accuracy was improved from the centimeter level to the millimeter level.

[0113] When the target value range is 500cm, the target is relatively close to the millimeter-wave radar. At this time, the electromagnetic wave signal is less affected by the delay of the radar's electronic components, distance delay, and meteorological factors. Therefore, and are relatively small, at 1.15mm and 1.75mm, respectively. However, as the target value range increases, that is, as the target distance from the millimeter-wave radar increases, the difference gradually increases, indicating that the accuracy of the correction value obtained by the linear fitting method gradually decreases. The maximum difference can reach 22.50mm, and except for the 500cm, 1300cm, 1700cm, and 2100cm value ranges, the difference in other value ranges is greater than 2mm. As shown in the last column, the difference of our method is generally well controlled at around 1mm, which is more stable than the linear fitting method. The millimeter-wave radar correction value obtained by our method is also more accurate.

[0114] Table 4 Schematic diagram of comparative experimental results

[0115]

[0116] in, It represents the error between the correction value of the linear fitting method and the true observation value. The calculation formula is: .

[0117] Through the above experiments and comparative analysis, it can be found that the predicted value of the present invention can well predict the ranging value of the millimeter wave radar after error elimination and mean processing, and its accuracy can be controlled at the millimeter level or even submillimeter level.

[0118] The implementation of each embodiment of the present invention is based on programmed processing performed by a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of each embodiment of the present invention are encapsulated into various modules. Based on this reality, and in addition to the aforementioned embodiments, an embodiment of the present invention provides a device for correcting distance measurements of millimeter-wave radars in tunnels that accounts for multiple sources of error. This device is used to perform the method for correcting distance measurements of millimeter-wave radars in tunnels that accounts for multiple sources of error, as described in the aforementioned method embodiments.

[0119] See also Figure 7 , the device includes:

[0120] Tunnel millimeter-wave radar ranging module, used to collect radar ranging values ​​and meteorological observation values;

[0121] The radar ranging value correction module is used to input the collected radar ranging values ​​and meteorological observation values ​​into the trained radar ranging value correction model and output the corrected millimeter-level tunnel radar ranging value; wherein the training of the radar ranging value correction model includes:

[0122] Obtain the results of the data collection experiment, including radar ranging values, meteorological observation values, and laser observation values, and construct a data set after preprocessing;

[0123] A radar ranging value correction model was constructed, consisting of a mixed LSTM and Bi-LSTM layer, a Dropout layer, and three fully connected layers. The connection structure was: LSTM and Bi-LSTM mixed layer—Dropout layer—first fully connected layer—second fully connected layer—third fully connected layer. The LSTM and Bi-LSTM mixed layer contained two LSTM layers and two Bi-LSTM layers, with a connection structure: first Bi-LSTM layer—first LSTM layer—second LSTM layer—second Bi-LSTM layer. Raw data was input into the mixed LSTM and Bi-LSTM layers, which then processed and output high-dimensional features. The high-dimensional features were then processed by the subsequent structure, and finally the corrected millimeter-level tunnel radar ranging value was output in the third fully connected layer.

[0124] Preset training parameters, input the radar ranging values ​​and meteorological observation values ​​in the dataset into the radar ranging value correction model, output the radar correction value, calculate the error between the radar correction value and the corresponding laser observation value in the dataset, adjust the model parameters through the back propagation algorithm, iterate the training, and output the trained radar ranging value correction model.

[0125] It should be noted that the device embodiments provided by the present invention are not only used to implement the methods in the above-mentioned method embodiments, but are also used to implement the methods in other method embodiments provided by the present invention. The only difference is the setting of corresponding functional modules. The principles thereof are basically the same as those of the above-mentioned device embodiments provided by the present invention. As long as those skilled in the art, based on the above-mentioned device embodiments, refer to the specific technical solutions in other method embodiments, obtain corresponding technical means and technical solutions composed of these technical means by combining technical features, and ensure the practicality of the technical solutions, they can improve the devices in the above-mentioned device embodiments to obtain corresponding device class embodiments, and obtain corresponding device class embodiments for implementing the methods in other method class embodiments. For example:

[0126] The tunnel millimeter-wave radar ranging system includes a millimeter-wave radar board and a meteorological parameter collection device. The millimeter-wave radar board sends millimeter waves to the radar target and receives the millimeter waves reflected by the target, and obtains the radar ranging value based on the time difference between sending and receiving. The meteorological parameter collection device collects meteorological observation values.

[0127] Furthermore, the correction equipment also includes a millimeter-wave radar board and a meteorological parameter collection device, which are respectively connected to the tunnel millimeter-wave radar ranging module, wherein the millimeter-wave radar board sends millimeter waves to the radar target and receives millimeter waves reflected by the target, and obtains the radar ranging value based on the time difference between sending and receiving, and the meteorological parameter collection device collects meteorological observation values.

[0128] Furthermore, it also includes: a laser tracker and a corner reflector, the corner reflector serves as a radar target for the millimeter wave radar board and the meteorological parameter acquisition device to respectively collect radar ranging values ​​and meteorological observation values, and the laser tracker is used to collect laser observation values.

[0129] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, located in one place or distributed across multiple network units. Some or all of these modules may be selected based on practical needs to achieve the objectives of this embodiment. Persons of ordinary skill in the art will understand and implement these embodiments without inventive effort.

[0130] The method of the embodiment of the present invention is implemented by electronic devices, so it is necessary to introduce the relevant electronic devices. Based on this purpose, the embodiment of the present invention provides an electronic device, such as Figure 6 As shown, the electronic device includes: at least one processor, a communications interface, at least one memory, and a communications bus, wherein the at least one processor, the communications interface, and the at least one memory communicate with each other via the communications bus. The at least one processor invokes logic instructions in the at least one memory to execute all or part of the steps of the methods provided in the aforementioned method embodiments.

[0131] In addition, when the logic instructions in the at least one memory are implemented in the form of a software functional unit and sold or used as an independent product, they are stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution is embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (a personal computer, a server, or a network device) to execute all or part of the steps of the method described in each method embodiment of the present invention. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, various media for storing program codes.

[0132] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0133] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0134] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0135] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0136] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A tunnel millimeter wave radar ranging value correction method taking into account multi-source errors, characterized in that: include: Collect radar ranging values ​​and meteorological observation values; The radar ranging value and the meteorological observation value are input into the trained radar ranging value correction model, and the corrected millimeter-level tunnel radar ranging value is output; The training of the radar ranging value correction model includes: Obtain the results of the data collection experiment, including radar ranging values, meteorological observation values, and laser observation values, and construct a data set after preprocessing; A radar ranging value correction model was constructed, consisting of a mixed LSTM and Bi-LSTM layer, a Dropout layer, and three fully connected layers. The connection structure was: LSTM and Bi-LSTM mixed layer—Dropout layer—first fully connected layer—second fully connected layer—third fully connected layer. The LSTM and Bi-LSTM mixed layer contained two LSTM layers and two Bi-LSTM layers, with a connection structure: first Bi-LSTM layer—first LSTM layer—second LSTM layer—second Bi-LSTM layer. Raw data was input into the mixed LSTM and Bi-LSTM layers, which then processed and output high-dimensional features. The high-dimensional features were then processed by the subsequent structure, and finally the corrected millimeter-level tunnel radar ranging value was output in the third fully connected layer. Preset training parameters, input the radar ranging values ​​and meteorological observation values ​​in the dataset into the radar ranging value correction model, output the radar correction value, calculate the error between the radar correction value and the corresponding laser observation value in the dataset, adjust the model parameters through the back propagation algorithm, iterate the training, and output the trained radar ranging value correction model.

2. The method for correcting the distance measurement value of a tunnel millimeter-wave radar taking into account multiple source errors according to claim 1, characterized in that: The radar ranging value is collected by the millimeter wave radar board, which sends millimeter waves to the radar target and receives the millimeter waves reflected by the target, and obtains the radar ranging value according to the time difference between sending and receiving; the meteorological observation value is collected by the meteorological parameter collection device, which is the measurement value of the environmental parameters in the tunnel, including but not limited to temperature, relative humidity, air pressure, wind speed, light pressure and PM 2.5 .

3. The method for correcting the distance measurement value of a tunnel millimeter-wave radar taking into account multiple source errors according to claim 1, characterized in that: The data collection experiment was designed as follows: Fix the millimeter-wave radar board and meteorological parameter collection device at the measurement point, fix the radar emission direction, install a laser tracker so that the laser emission direction and distance are consistent with the direction and distance of the radar, set the collection distance range, and divide the collection distance range into target value segments of equal distance. The corner reflector is used as the radar target and installed in the radar emission direction. The distance from the millimeter-wave radar board is changed according to different target value segments. For each target value segment, experimental data is collected within the preset fixed collection time.

4. The method for correcting the distance measurement value of a tunnel millimeter-wave radar taking into account multiple source errors according to claim 1, characterized in that: The experimental data were collected using a time-divided repeated observation scheme, which divided the preset fixed collection time into time periods. For each target value segment, the experimental data were repeatedly collected in different time periods. During the experiment, a laser tracker was used to collect laser observation values ​​from the measurement point to the millimeter-wave radar target.

5. The method for correcting the distance measurement value of a tunnel millimeter-wave radar taking into account multiple source errors according to claim 1, characterized in that: The working steps of the radar ranging value correction model are specifically as follows: The first Bi-LSTM layer captures the long-term dependencies between radar ranging values ​​and contextual information about meteorological parameters in the original data sequence. The first LSTM layer uses memory cells and gating mechanisms to prevent overfitting and learn to preserve long-term dependencies. The second LSTM layer deeply extracts sequence features, further learning and capturing the complex patterns of long-term dependencies between sequences. Finally, the second Bi-LSTM layer receives the results of the second LSTM layer and integrates forward and backward information to obtain high-dimensional features with rich feature representation. The results of the second Bi-LSTM layer are input to the Dropout layer. The dropout rate is set to randomly discard the output of the Dropout layer neurons. The high-dimensional features after the discarding process are linearly transformed and fused by the first fully connected layer. Then, the second fully connected layer reduces the dimension and fuses the features. Finally, the third fully connected layer maps them to a one-dimensional regression prediction value as the corrected millimeter-level tunnel radar ranging value.

6. A tunnel millimeter wave radar ranging value correction device taking into account multi-source errors, characterized in that: include: Tunnel millimeter-wave radar ranging module, used to collect radar ranging values ​​and meteorological observation values; The radar ranging value correction module is used to input the collected radar ranging values ​​and meteorological observation values ​​into the trained radar ranging value correction model and output the corrected millimeter-level tunnel radar ranging value; wherein the training of the radar ranging value correction model includes: Obtain the results of the data collection experiment, including radar ranging values, meteorological observation values, and laser observation values, and construct a data set after preprocessing; A radar ranging value correction model was constructed, consisting of a mixed LSTM and Bi-LSTM layer, a Dropout layer, and three fully connected layers. The connection structure was: LSTM and Bi-LSTM mixed layer—Dropout layer—first fully connected layer—second fully connected layer—third fully connected layer. The LSTM and Bi-LSTM mixed layer contained two LSTM layers and two Bi-LSTM layers, with a connection structure: first Bi-LSTM layer—first LSTM layer—second LSTM layer—second Bi-LSTM layer. Raw data was input into the mixed LSTM and Bi-LSTM layers, which then processed and output high-dimensional features. The high-dimensional features were then processed by the subsequent structure, and finally the corrected millimeter-level tunnel radar ranging value was output in the third fully connected layer. Preset training parameters, input the radar ranging values ​​and meteorological observation values ​​in the dataset into the radar ranging value correction model, output the radar correction value, calculate the error between the radar correction value and the corresponding laser observation value in the dataset, adjust the model parameters through the back propagation algorithm, iterate the training, and output the trained radar ranging value correction model.

7. The tunnel millimeter wave radar ranging value correction device taking into account multi-source errors as claimed in claim 6, characterized in that: The correction equipment also includes a millimeter-wave radar board and a meteorological parameter collection device, which are respectively connected to the tunnel millimeter-wave radar ranging module. The millimeter-wave radar board sends millimeter waves to the radar target and receives millimeter waves reflected by the target, and obtains the radar ranging value based on the time difference between sending and receiving. The meteorological parameter collection device collects meteorological observation values.

8. The tunnel millimeter wave radar ranging value correction device taking into account multi-source errors as claimed in claim 7, characterized in that: Also includes: A laser tracker and a corner reflector, wherein the corner reflector serves as a radar target for the millimeter wave radar board and the meteorological parameter acquisition device to respectively collect radar ranging values ​​and meteorological observation values, and the laser tracker is used to collect laser observation values.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium stores computer instructions, which cause the computer to execute the method of any one of claims 1 to 7.

Images